Pre-auth RCE in nginx is the bug every hacker dreams of finding. It almost never happens.
Today we’re disclosing nginx2root: five nginx vulnerabilities, and an exploit that takes an unauthenticated attacker all the way to root.
One of themhits the default config.
If you run nginx, patch now. Then read: nginx2root.com
CVE-2026-123007, CVE-2026-123634, CVE-2026-123633, CVE-2026-121810, CVE-2026-113425
Software can be verified and mathematically perfect down to the bit but every proof relies on assumptions made about the hardware running the software and the environment it runs in.
@AntithesisHQ wrote a great blog post about finding bugs in formally verified Raft implementations due to this exact mismatch.
Cancellation is really hard to get right. While studying tokio-uring I found a use-after-free you can reach from safe Rust.
With io-uring the kernel holds a pointer to your buffer until it posts the op’s final completion. Even if the future is dropped the runtime must ensure the buffer stays alive until then.
During shutdown tokio-uring has a bug where it:
1) drops the buffer for an op who’s future has been dropped
2) then sends a cancel event to the kernel
With this order of operations if data arrives the kernel can write to the buffer even after the runtime has dropped it but before it has sent the cancellation event.
I've been getting super into how different Rust io_uring runtimes manage buffer ownership.
Unlike with epoll, an io_uring operation can outlive the future that submitted it, leading to some interesting challenges.
Currently digging into tokio-uring's approach.
This is literally the point of rust! I can now give extra scrutiny to those 4832 unsafe blocks and know that the compiler will enforce memory safety everywhere else.
Meanwhile in C the whole codebase is effectively a giant unsafe block
Someone re-wrote curl in Rust (because rewriting working code is required by the Cult of Rust).
That Rust curl clone contains 4,832 uses of “unsafe” in order to make it work.
Because the point of Rust is MeMoRY sAfEtY. 🤪
Why stop at tests? I personally put it in my AGENTS.md to skip the source code entirely and just patch the production binary directly.
Programming languages are just an abstraction invented by us stupid apes because we can’t keep all the logic in our heads.
i deleted all the tests in my repo and banned them in AGENTS.md
tests were always just a second, worse copy of your logic. a leaky imitation humans wrote because we couldn't hold every invariant in our head (staying green half the time the real bug happened anyway)
opus 5.5+ is good enough at raw logic now to read the intended invariants straight from the codebase
maintaining that duplicate copy is just economic waste. plus it makes tasks take way longer while agents burn compute running tests when they almost certainly got it right anyway